A NEW ALGORITHM FOR SELF-ADAPTING

WEB INTERFACES

Bogdan Vintila, Dragos Palaghita and Maria Dascalu

Economic Informatics Department, University of Economics Bucharest, Calea Dorobanti Street, Bucharest, Romania

Keywords: Self-adapting web interface, Software development, Interface design, Dynamic web interfaces, Statistical

analyses.

Abstract: The paper proposes a new improved algorithm for creating hierarchies of features and options for self-

adapting web interfaces against the common one used by many applications. The user interface concept is

presented. Types of user interfaces are described. Quality characteristics of the user interfaces are analyzed.

Ways of fulfilling these quality characteristics while keeping the costs low are discussed. Advantages and

disadvantages of self-adapting and static web interfaces are given. The most common algorithm for creating

hierarchies of features and options is described and analyzed. The advantages and disadvantages of the

proposed algorithm are discussed. New directions for the development of the self-adapting web interfaces

are highlighted.

1 INTRODUCTION

User interfaces are pieces of software that ensure the

interaction between the program’s logic and the

user. In order to be successfully utilized by users, a

software product needs not only to have flawless

functionality but also to expose its features in an

accessible way. Even if the program has rich

features that work perfectly, a deficient user

interface leads to the fail on market. This is

explained by the users’ lack of time for learning how

to use new software and thus they orient towards

easy to use software. User interfaces experienced a

great evolution from the beginnings of the

computing age. As user interfaces are used for the

communication between the user and the application

they must facilitate the selection of processing

options, data input and output. The user interfaces

started with the batch interface in 1945 (Wikipedia,

2009). In 1969 the batch interface was replaced by

the command-line user interface. The command-line

user interface has the advantage of giving the user

access to all commands and parameters. The main

disadvantage is that the user must know the syntax

of all options and commands he wants to access.

This led to the apparition of the Graphical User

Interface (GUI) in 1981. Nowadays the command-

line user interface addresses highly trained

professionals that need precise control and access

without the performance overhead of the graphical

interface.

There are many types of user interfaces

(Wikipedia, 2009):

- Graphical User Interfaces;

- Web User Interfaces; are used by web

applications; the user utilizes a Internet

browser to access the application; data input is

made through a form generated by the server;

results’ visualization is made also through web

pages generated by the server and loaded by

the users’ Internet browser (Hall Mary W.,

2008); nowadays these are widely used as web

applications have many advantages against

standalone ones;

‐ Command-line Interfaces;

‐ Touch User Interfaces;

‐ Gesture Interface;

‐ Multi-screen Interfaces;

‐ Motion Tracking Interfaces;

‐ Voice User Interfaces.

Standalone applications are difficult to update.

The process depends on the user, so compatibility to

earlier versions must be ensured. Web applications,

on the other side, are easy updateable as the

application resides on one or more physical servers.

Also, web applications can be accessed regardless of

the location or machine. Presently, web applications

Vintila B., Palaghita D. and Dascalu M.

A NEW ALGORITHM FOR SELF-ADAPTING WEB INTERFACES.

DOI: 10.5220/0002773400570062

In Proceedings of the 6th International Conference on Web Information Systems and Technology (WEBIST 2010), page

ISBN: 978-989-674-025-2

tend to replace standalone applications in more and

more domains. Given the importance of these

applications, their interfaces are of high importance

as they ensure the communication between the user

and the application’s logic.

In (Esko Juuso, 1997) the role of neural

networks, genetic algorithms and linguistic

equations in the development of user interfaces are

described. Many challenges in the development of

adaptive interfaces and also many research

directions are given in (Langley, 1999).

Adapting the interface to the needs of the user

reduces its complexity and improves efficiency.

Awareness is analyzed as an evaluation metric in

order to obtain the reduced visual complexity of

interfaces (Findlater & McGrenere, 2009).

The study presented in (Letsu-Dake & Ntuen,

2009) shows clearly that the adaptive interfaces are

better in terms of time to achieve the goal under

fault conditions, fault detection time, number of

system failures and fault detection rate.

In (Pietschmann et al., 2009) a study of the

business processes is made through the eye of user

interfaces. User interface services provide many

rich, reusable components for building user

interfaces.

The importance of the adaptive interfaces for

special domains such as the automotive industry is

highlighted in (Amditis et al., 2006). A

methodological framework for optimizing the

human machine interfaces is presented and the

results of its implementation are discussed.

The concept of Universal Access is presented by

(Stephanidis, 2001) and also ways of building

systems that ensure interfaces capable of adapting

according to the context. Universal access is attained

through the adaptive interfaces.

The importance of the adaptive interfaces is

analyzed in (Savidis & Stephanidis, 2004) for a wide

range of devices. The unified user interfaces

described by the paper are capable to self-adapt at

run-time according to the needs and requirements of

the current user and device. The development

process of unified user interfaces is also described

by the authors.

A methodological approach for modeling

adaptation decisions and for solving the problem of

integrating existing as well as acquired knowledge in

the decision module of an adaptive interface is

proposed in (Zarikas, 2007). The model uses

influence diagrams and it provides a method of

encoding user and context information as well as

other factors that are involved in the decision

making process. The paper also presents an

illustrative example of the analyzed modeling

method.

The notions of user profile, interface profile and

the compound usability are discussed in (Nguyen &

Sobecki, 2003). Their role in the development of the

adaptive interfaces is also analyzed. Using

consensus-based methods, the authors build

interface profiles appropriate to classes of users.

2 QUALITY CHARACTERISTICS

OF USER INTERFACES

All interfaces must fulfill quality requirements in

order to give the users the maximum degree of

satisfaction. Some of the most important quality

characteristics are (Ivan et al., 2008):

‐ clarity;

‐ conciseness;

‐ familiarity;

‐ responsiveness;

‐ consistency;

‐ attractiveness;

‐ efficiency;

‐ forgiveness.

These characteristics, if fulfilled at the same

time, ensure the success of the user interface.

Anyway, by increasing one of the characteristics it is

possible to lower others so achieving a balance

between then is a complex and long process (Julio

Abascal, 2008). Considering the application’s target

group it is possible to satisfy some requirements

with lower levels of some characteristics leaving

additional time to increase the level of more

important ones.

The measurement of the quality characteristics is

usually a hard task as there are no clear procedures

of measurement for each of them. Let us consider

responsiveness. For this characteristic the obtained

value upon measurement depends on the user if no

measurement procedure is stated. If the application

is unresponsive for one second after the user gives a

command, the user might consider it lacks a good

response time or might even not notice the delay. If

a measurement procedure is defined there is no room

for subjective judgment. If the procedure states an

interface is responsive if it has delays less than 1

second, the considered case lacks responsiveness. If

the procedure states a time of more than 2 seconds,

the considered case is responsive.

WEBIST 2010 - 6th International Conference on Web Information Systems and Technologies

Table 1: Advantages and disadvantages of self-adapting and static web interfaces.

Self-adapting web interfaces Level of the user Static web interfaces Level of the user

Advantages Clear

Light

Efficient

Attractive

Simple

Easy to use

Change depending on

context

Change in time

Low bandwidth usage

Suggest additional features

End-user

Efficient

Detailed

Doesn’t change in time or

according to context

Allow workflow

automation

Detailed control

Professional

Disadvantages Don’t allow detailed control

Force users to request

additional features

Professional

Professional and End-

user

Overcrowded

High complexity

Can’t be efficiently used

Unattractive

High bandwidth usage

End-user

3 STATIC VS. SELF-ADAPTING

WEB INTERFACES

The pass from static user interfaces to the self-

adapting ones was done in order to facilitate the use

of software by untrained users and also increase the

efficiency. Self-adapting interfaces also decrease the

operating time by displaying the right features in the

right spot and hiding the unnecessary ones (Ion Ivan,

2009). Self-adapting user interfaces are based on

studies regarding the users’ comportment in a certain

application or when trying to solve a certain

problem. These have a predictable comportment as

the frequency indicators don’t change during

application’s use or only in a small proportion.

There are also interfaces that study the user’s

comportment and are dynamically built based on the

indicators calculated using these data. E-Commerce

sites suggest products based on user’s comportment

or similarities with other users (Sharifi & all, 2004).

Menu options hide until the first use. Options are

showed in the order they are used most.

Static user interfaces are used for applications

that have few features and options or for ones that

address highly trained professionals. Static

interfaces are difficult to use for end-users, but

professionals prefer them as they can automate

workflow. The self-adapting web interfaces have the

advantage of conserving precious bandwidth. By

determining the most used features of the

application, the graphical elements corresponding to

the other features don’t usually have to be loaded at

the start. If necessary, the user can request the load

of additional components as he uses the application.

The lower the used bandwidth is, the higher the

user’s experience gets.

Advantages and disadvantages of the self-adapting

and static web interfaces are given in (Table 1).

As seen in (Table 1) advantages and disadvantages

are taken into account considering the interface’s

target group.

Detailed control is a must be for the highly

trained professional, but for the end-user is not

important. Also the interface’s ability to change in

time is an advantage for the end-user but a

disadvantage for the professional that quickly learns

the features’ options and location. As users differ in

training level, the same interface has positive or

negative impact in use. The higher the training level,

the more interfaces the user can use and overcome

their disadvantages. Both types of interfaces have

advantages and disadvantages so the users must

choose the one that makes him more efficient.

4 COMMON ALGORITHM

Self-adaptive web interfaces must adapt their

behavior from an application’s run to another and

even during the work session (Savidis &

Stephanidis, 2004). By recording and analyzing

user’s comportment it is possible to obtain such

results. There are many algorithms to predict user’s

comportment based on past actions (Gena C., 2007).

The most common assumes the following:

Let F be a set of n features the user can access in

a given application A. An application’s feature is

defined by a succession of steps that, through the

processing of the input data, lead to wanted results.

A NEW ALGORITHM FOR SELF-ADAPTING WEB INTERFACES

For example, the print process that is present in

many software products, is a feature of them.













,...,







(1)

Let Q be a set of n frequencies associated to the

set of features. Feature one, f

is associated to q

, f

to q

and so on.





,



,…,





(2)

On the base on Q set a hierarchy of the features

is built. The most accessed features rank the highest

(Ivan et al., 2009).

Let us consider the A application was never used and

all frequencies are 0. Considering this, all features

have the same priority. At the first use, feature 



used two times, 



is used once and 



is used

once.

At the second use, the Q set has the following

values:

  2,0,1,0, … ,0,1,0

(3)

These lead to the following hierarchy:



























,…,











(4)

Considering H, at the second run 



is the first

one in the selection list, 



is the second, 



is the

third and the ones with 0 frequency follow. This

way, the user is more likely to have the feature he

wants to access at the top of the selection list. The

efficiency of the algorithm increases with every use

of the application.

The possibility of having individual interfaces

for each user, rather than the whole application, is

given by storing data for each user and building the

interface based on that data (Brusilovsky, 2001).

5 THE IMPROVED ALGORITHM

A new algorithm is proposed to improve user

experience. The algorithm is based on calculating

scores and ranking features on the bases of

dependencies between features.

Let 



,



 be the function that returns the

dependency value between features i and j. The

highest the value of 



,



 is, the higher is the

probability that the user selects feature j after

selecting feature i. A dependency table is obtained

by calculating the D function for all features. The

main diagonal elements are equal to 0 as the

selection of a feature can’t determine the repeated

selection leading to an infinite loop. When a feature

is accessed by the user, the score of the features that

depend by this one are updated accordingly to the

dependency’s strength. This leads to better scores

for features that were not selected, but have a very

high probability of being selected.

Let 







,





0.6, 







,





0.2 and all the

other dependencies be 0.

Taking into consideration this system of ranking,

the above dependencies and the number of feature

access above, the resulting Q set is





 2,1.2,1,0.24,0, … ,0,1,0

(5)

leading to the following 



hierarchy































,…,











(6)

Comparing H to H

we see that the order of

features is different, 



being the second feature in

the list before 



and 



. The interface updates

according to this hierarchy either at set moments,

either when the changes that arise between the old

arrangement and the new one grows above a set

value. Using the dependencies system better

forecasts of the user’s comportment can be made,

but only if the dependencies are determined using

very large datasets for analyses. The use of incorrect

dependencies between features results in incorrect

selection lists. These cause poor user experience as

instead of quickly solve its problem, the user tries to

find the desired feature. As the user utilizes more

and more the application, the frequency set Q gets to

be more and more significant for the user’s

comportment and the application will make better

forecasts regarding future user’s actions.

A very important issue is choosing the D

function that returns the dependencies between

values. A very simple, yet effective, way of

establishing dependencies is by analyzing users’

behavior in existing applications. Let us consider an

e-commerce application is to be built. For this we

can analyze the data on users’ behavior that other e-

commerce applications already have. After filtering

and clustering the data, dependencies of the clusters

are determined and these are used for the

dependencies set of features (Lau & Horvitz, 1999).

Even if the e-commerce is borderless, the percentage

of local users is greater than the percentage of global

users. This can lead to good predictions on new

users’ behavior only for local ones. This can be

WEBIST 2010 - 6th International Conference on Web Information Systems and Technologies

avoided by studying data from many applications

worldwide.

Compared to the basic algorithm, this one

improves the success rate of having a desired feature

closer to the user.

6 ALGORITHM VALIDATION

A sample application is being currently developed to

exemplify the self-adapting web interface using the

dependency enabled rank computing algorithm.

As the algorithm states, the ranking system is

based on the addition to the base frequency of the

dependency score. The dependencies used are not

bidirectional thus 



,







,



. This is

normal because action a determines action b, but

action b can’t determine action a. The application is

to implement real functionality and is to be released

to the public to test the algorithm’s validity. The

time needed by users to select options is to be

recorded and the obtained data is to be analyzed

leading to the algorithm’s validity or invalidity.

Short times between the selections of features

indicate a good prediction of the user’s

comportment. Long times for feature selection

indicate the algorithm’s failure in providing

qualitative forecasts. Collected data must be

preprocessed as to include in the analysis only data

obtained by users after several use of the application

when they understand the application’s adaptive

comportment.

7 CONCLUSIONS

With time many types of interfaces were developed.

The diversity is given by advantages of each type of

interface for a certain type of application or process.

Self-adapting web interfaces are of great future as

more and more users have low training. These

interfaces allow the untrained users to use the

informational systems at a basic and advanced level

without training. The common algorithm for creating

hierarchies can be improved by taking into account

dependencies between the collectivity’s elements.

The self-adaptive interfaces are of great importance

in domains such as: e-learning, e-governance, office

suites, operating systems, mobile applications.

Future research includes color coding the

features so that the color best perceived by the

human eye is associated with the feature most

probable the user will access, the second color in the

perception hierarchy is associated to the second most

probable feature and so on. Future research also

aims at the use of cameras to keep the user’s eyes

under observation and detect the screen zones the

user focuses most and thus placing there the most

accessed features and options. By detecting the

direction of human gaze, the navigation within the

interface is possible. The user focuses the desired

option and based on his position, distance from the

camera and previous configuring, the application

detects the selected area and activates the feature.

ACKNOWLEDGEMENTS

This article is a result of the project „Doctoral

Program and PhD Students in the education research

and innovation triangle”. This project is co funded

by European Social Fund through The Sectorial

Operational Programme for Human Resources

Development 2007-2013, coordinated by The

Bucharest Academy of Economic Studies.

REFERENCES

Amditis, A., Polychronopoulos, A. & Andreone, L., 2006.

Communication and interaction strategies in

automotive adaptive interfaces. Cognition, Technology

and Work, 8(3), pp.193-99.

Brusilovsky, P., 2001. Adaptive Hypermedia. User

Modeling and User-Adapted Interaction, pp.87-110.

Esko Juuso, K.L., 1997. Adaptive Interfaces and Soft

Computing. Studies in Informatics and Control.

Findlater, L. & McGrenere, J., 2009. Beyond

performance: Feature awareness in personalized

interfaces. International Journal of Human-Computer

Studies, 68(3), pp.121-37.

Gena C., W.S., 2007. Usability Engineering for the

Adaptive Web. In The Adaptive Web. Springer Berlin /

Heidelberg. pp.720-62.

HALL Mary W., G.Y.L.R., 2008. Self-Configuring

Applications for Heterogeneous Systems: Program

Composition Using Cognitive Techniques.

Proceedings of the IEEE, 96(5), pp.849-62.

Ion IVAN, B.V.C.C.M.D., 2009. The Modern

Development Cycle of Citizen Oriented Applications.

Studies in Informatics and Control, 18(3).

Ivan, I., Vintila, B. & Palaghita, D., 2008. Quality metrics

of citizen oriented informatics applciations. In Forth

International Conference on Applied Statistics.

Bucharest, Romania, 2008.

Ivan, I. et al., 2009. Collectivity's elements ranking. In

16th International Economic COnference IECS 2009.

Sibiu, Romania, 2009.

A NEW ALGORITHM FOR SELF-ADAPTING WEB INTERFACES

Julio Abascal, I.F.d.C.A.L.J.M.C., 2008. Adaptive

Interfaces for Supportive Ambient Intelligent

Environments. In ICCHP 2008., 2008. Springer Berlin

/ Heidelberg.

Langley, P., 1999. User modeling in adaptive interfaces.

In Proceedings of the seventh international conference

on User modeling. Banff, Canada, 1999. Springer-

Verlag New York, Inc. Secaucus, NJ, USA.

Lau, T. & Horvitz, E., 1999. Patterns of search: analyzing

and modeling Web query refinement. In Proceedings

of the seventh international conference on User

modeling. Banff, Canada, 1999. Springer-Verlag New

York, Inc. Secaucus, NJ, USA.

Letsu-Dake, E. & Ntuen, C.A., 2009. A case study of

experimental evaluation of adaptive interfaces.

International Journal of Industrial Ergonomics, 40(1),

pp.34-40.

Nguyen, N.T. & Sobecki, J., 2003. Using consensus

methods to construct adaptive interfaces in multimodal

web-based systems. Universal Access in the

Information Society, 2(4), pp.342-58.

Pietschmann, S., Voigt, M. & Meibner, K., 2009.

Adaptive rich user interfaces for human interaction in

business processes. In Lecture Notes in Computer

Science., 2009.

Savidis, A. & Stephanidis, C., 2004. Unified user interface

development: the software engineering of universally

accessible interactions. Universal Access in the

Information Society, 3(3-4), pp.165-93.

Sharifi, G. & all, 2004. Location-Aware Adaptive

Interfaces for Information Access with Handheld

Computers. In Adaptive Hypermedia and Adaptive

Web-Based Systems. Springer Berlin / Heidelberg.

pp.328-31.

Stephanidis, C., 2001. Adaptive Techniques for Universal

Access. User Modeling and User-Adapted Interaction,

11(1-2), pp.159-79.

Wikipedia, 2009. User Interface. [Online] Available at:

http://en.wikipedia.org/wiki/User_interface [Accessed

16 October 2009].

Zarikas, V., 2007. Modeling decisions under uncertainty

in adaptive user interfaces. Universal Access in the

Information Society, 6(1), pp.87-101.

WEBIST 2010 - 6th International Conference on Web Information Systems and Technologies